Re: [PATCH 1/3] ata: sata_dwc_460ex: use "dmas" DT property to find dma channel

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Mon, 2015-12-21 at 19:27 +0000, Måns Rullgård wrote:
> Andy Shevchenko <andriy.shevchenko@linux.intel.com> writes:
> 
> > On Mon, 2015-12-21 at 01:19 +0000, Måns Rullgård wrote:
> > > Andy Shevchenko <andy.shevchenko@gmail.com> writes:
> > >  
> > > > P.S. I also noticed that original driver enables interrupt per
> > > > each
> > > > block
> > > 
> > > And then ignores all but the transfer complete interrupt.
> > > 
> > > > and sets protection control bits.
> > > 
> > > With no indication what the value it sets is supposed to mean.
> > 
> > Okay, let's summarize what we have:
> > 
> > 0. AR: Get a working reference for PPC 460EX SATA driver
> 
> Do we consider Julian's latest result working?

I think so.

> 
> > 1. AR: Clear LLP_EN bits at the last block of LLP transfer
> 
> Patch sent.

Acked.

> 
> > 2. AR: Rename masters to 'memory' and 'peripheral' and change them
> > per
> > DMA direction
> 
> Good idea.  I'd call them memory and device though to match existing
> dmaengine nomenclature.

I remember how I called in my patch. So, there is no problem to rename,
but will see.

> 
> > 3. AR: Set LMS (LLP master) to 'memory' when do LLP transfers
> 
> I started working on a patch for that already.

Thanks.

> 
> > 4. CHECK: PROTCTL bit (documentation says that recommended value is
> > 0x01)
> 
> Any idea what the value of 0x3 used by the old sata driver means?
> Presumably that's decided by the bus.

Nope, documentation says that it is direct representation of hprot[3:1]
wires on the master interface. Also it refers to AMBA spec, so, if you
have access to AMBA spec I think we might get it from there.

> 
> > 5. CHECK: Other bits in CFG register (FIFO_MODE, FCMODE)
> > 6. CHECK: Block interrupts vs. one interrupt at the end of block
> > chain
> > (Måns, I missed how any of them is ignored)
> 
> The interrupt handler looks at the StatusTfr and StatusErr registers
> and
> ignores StatusBlock.

I have to refresh my memory, since BLOCK interrupts should be enabled
(unmasked) separately. I have forgotten which type of interrupt is
generated in this case, BLOCK, or XFER after each block, or only one
XFER at the last block (LLP.LOC = 0) and BLOCK are ignored. So, will
check later.

> 
> > 7. AR: Test everything on Intel SoCs such as Baytrail, CherryTrail,
> > etc
> > (SPI, UART, dmatest), AVR32 (MMC, dmatest), PPC 460EX (Onboard
> > SATA)
> 
> I can test on AVR32.  That is as far as I know the only system I have
> with this DMA engine.

If you have Intel Haswell, BayTrail, Braswell, CherryTrail, Broadwell,
you have it as well as long you have LPSS block there. (Most of them
are Atoms).

> 
> > I can share my working branch with a set of patches regarding to
> > dw_dmac. We may do our work based on that code and after I'll
> > submit
> > everything to upstream. Does it sound okay for you, guys?
> 
> I'm going away for the holidays, so I won't be able to do any serious
> work on this until January, but I'll keep an eye on emails and may
> even
> reply occasionally.  Before I go, I'll publish my patches so far
> whatever shape they're in.

Okay, thanks! I will include them in my branch which I'm going to
publish on GitHUB.

Happy holidays!

-- 
Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Intel Finland Oy